This document discusses the work of three roboticists at GW:
1) Gabe Sibley, who develops perception algorithms for robots like Boston Dynamics' LS3 robot and autonomous vehicles. His lab builds agile "ninja cars" to test these algorithms.
2) Pinhas Ben-Tzvi works on perception and control for robots that can operate in degraded conditions like night or snow.
3) They aim to develop robots that can walk with soldiers, rescue victims, and assist around the home. Autonomous vehicles will change transportation by allowing tasks like napping or working while commuting.
This document discusses the work of three roboticists at GW:
1) Gabe Sibley, who develops perception algorithms for robots like Boston Dynamics' LS3 robot and autonomous vehicles. His lab builds agile "ninja cars" to test these algorithms.
2) Pinhas Ben-Tzvi works on perception and control for robots that can operate in degraded conditions like night or snow.
3) They aim to develop robots that can walk with soldiers, rescue victims, and assist around the home. Autonomous vehicles will change transportation by allowing tasks like napping or working while commuting.
This document discusses the work of three roboticists at GW:
1) Gabe Sibley, who develops perception algorithms for robots like Boston Dynamics' LS3 robot and autonomous vehicles. His lab builds agile "ninja cars" to test these algorithms.
2) Pinhas Ben-Tzvi works on perception and control for robots that can operate in degraded conditions like night or snow.
3) They aim to develop robots that can walk with soldiers, rescue victims, and assist around the home. Autonomous vehicles will change transportation by allowing tasks like napping or working while commuting.
The Body Robotic Form, function, and the future as seen through the eyes and handiwork of three GW roboticists. By Danny Freedman | photography by Jessica McConnell Burt One of them aspires to be on the battlefield. Another, in the home as a domestic aide. Others aim to perform surgeries and to chauffeur the tired, the busy, and the infirm. They are unseasoned immigrants to these parts of the work- force, though, and are about as green as they come. Numb to social cues and the tug of common sense, they bring only a heat- seeking determination. “In the last couple decades robots started roaming out of the industrial corridor,” says Pinhas Ben-Tzvi, an assistant professor of mechanical and aerospace engineering. “They’re swimming, they are flying, they are walking.” But in order for robots to fully merge into society’s fast lane, as they are being coaxed to do, they will need a depth that is almost more than the sum of their parts: They will need to perceive the world around them and make complicated decisions that power sophisticated maneuvering. And they’ll need to do it all inside the tangled and fragile mess called daily life. Here, three modern-day Geppettos working on the future of robotic perception, control, and mechanics open their workshops and share visions of a new generation of robots: ones that will walk alongside U.S. soldiers, pull victims from rubble, and lend a hand—or something like a hand—around the house. To test perception algorithms being designed for high-profile robots like Boston Dynamics' AlphaDog and an autonomous car by Toyota, Dr. Sibley's lab has built five "gymnastic, parkour-style" robots, including this one, nicknamed Herbie.
G W R E S E A R C H • fa l l 2 0 1 2 19
40726_GWU_018-023.indd 19 4/16/13 10:38 AM
My work focuses on that fast and accu- right angle, so it can hit the next jump in rate perception of the world, as well as order to go off the half-pipe at just the the high-speed planning and control. It’s right velocity and not crash. called perception-driven model-predictive- control, and we use it for fast and agile ground vehicles that jump, slide, and bounce over complicated 3D terrain. A utonomous cars are going to completely change the way we relate to the automobile. When people are driving about 55 miles per hour, only
F or Toyota’s autonomous cars we’re
working on robust perception, plan- ning, and control algorithms capable of 12 percent of the highway is occupied by a vehicle—that’s because people don’t want to get too close to each other. negotiating degraded conditions: driving at Because it’s scary. We’re really bad at night, driving in the rain, or snow. driving in these machines. And we can Cars that drive themselves will need build autonomous systems that don’t to understand the road in difficult situa- have to look in mirrors to see backwards. Herbie and the other "ninja cars" are built from tions. It’s not always going to be a sunny They’re looking 360 degrees all the time. off-the-shelf remote-controlled cars, which over drive in California—sometimes it’s going Fifty years from now our children will the years, Dr. Sibley says, have become rugged, to be sliding sideways at night in Sweden. say: Wait a second—did you really get in those light, and inexpensive. In these radical situations we still want to metal boxes and push and pull levers and gears come up with the right answer. To do that, to make them move? Wasn’t that dangerous? Gabe Sibley our goal is to develop perception, planning, Why didn’t you have the computer do it? And and control systems that supersede even they’ll be right. When you grab the wheel Department of Computer Science Mario Andretti’s. your insurance premium will just go up. Sample projects: robotic perception We’re doing essentially the same Of course, there are all sorts of fun soci- for Boston Dynamics’ LS3 robot and thing for platforms like LS3, nicknamed ological issues, legal issues, and policy issues. for an autonomous car by Toyota AlphaDog. [More formally, that is Boston There’s just no practical way to go from Dynamics’ Legged Squad Support System, kind-of autonomous— where we are now, We look for algorithms that are the a four-legged, all-terrain pack mule, being with systems for things like lane-keeping underpinning of perception and action built for the military, that will be capable and parking—to fully autonomous. What and use them to build autonomous of traveling autonomously, either alone or we have to do is make cars that are capable machines that can perform useful tasks. trailing soldiers.] of driving in human traffic. Using cameras and computer vision we It’s a very similar problem. Robust When you commute you can take a nap try to teach robots how to “see”—how perception is so hard from a platform that or maybe do your email. If you’re impaired, to understand the spatial and semantic can bound through very rough terrain at you don’t have to be chauffeured around. context they share with us. So that’s: excessive speeds, making it difficult to see It’s going to be more efficient. It’s going to Where am I? Where and what are objects of what is going on from the on-board camera. be better for the environment. Autonomous interest here? So we have to take that very blurry, noisy highways are going to revolutionize society. We have to compute these solutions image data, inertial data, and other sensor quickly, fast enough so that, for example, data and make sense of them to build accu- if I’m a robot running through the woods rate internal world models. Pinhas Ben-Tzvi I can avoid hitting trees and stepping in As a test platform for these algorithms the wrong place. we have these small, fast, and robust ground Department of Mechanical & To do that the robot has to build a vehicles that can handle high-speed jumps. Aerospace Engineering mental picture of the world, like an internal Imagine putting a small robot inside a skate Sample projects: mobile robotics, Pixar movie. That’s then used to make park and turning it loose. including work for the military’s decisions—specifying an action and simu- The students call that project “ninja Defense Advanced Research Projects lating the physics to predict the result. car” but it’s really about the perception, Agency (DARPA), and robotic surgery It’s planning over that mental model, the modeling and tracking of what it’s like a gymnast visualizing the flip and seeing. And much more, even, it’s about the My lab conducts fundamental and tumble before they do it; imagining planning side of that: coming up with the applied research in robotics and mecha- what would happen and using the choices to make in order to hit that jump, tronics (the synergy of mechanics, results to execute the move. and to land at the right spot, at exactly the electronics, and computer control in an
20 GW RESEARCH • Spring 2013
40726_GWU_018-023.indd 20 4/16/13 10:38 AM
Dr. Ben-Tzvi’s novel hybrid mechanism allows both this robot’s arm and traction mechanism to be dual-purpose, with either capable of aiding movement or manipulation. A pop-up vision system and a gripper (both visible at right) retract into the arm, enabling the robot to be flat and flippable. Front and back cameras, lights, and sonar are embedded near the tracks to guide a remote operator.
integrated design). The beauty of trans-
forming that fundamental research into different applications is what drives me. For instance, we are working on devel- oping mobile robotic systems for search- and-rescue applications, for reconnaissance, for inspection, for monitoring; they could also be used for handling improvised explo- sive devices and bomb disposal. We are also working on medical applications, like robotic systems for surgery. The goal is to benefit from robots, to have a better, safer, healthier, and easier life.
I was motivated by 9/11. I read a lot of
articles about how robots were able to do some useful tasks but weren’t dexterous enough or robust enough to reach deeper into the rubble. The prevailing design of mobile robots is based on a traction mechanism for locomotion, with a separate arm attached on top for manipulation. But if a robot like that is used on rough terrain and it flips, the arm could break. So I came up with this idea that I call the hybrid mechanism. It has links that can be used interchangeably for locomotion and manipulation—and also do both simultaneously. It gives the system far greater capa- bilities. For example, if you’re trying to lift something, imagine being able to use your leg as an arm. You’d be able to lift a lot more. The robot has tracks on the bottom so it can travel on rough terrain, and the arm can be used as leverage for things like climbing obstacles, going up and down stairs or over ditches and other terrains. An integrated hand can be used to open doors. The robot is fully symmetrical when the arm is folded down, so it can flip and still continue operating.
GW RESEARCH • Spring 2013 21
40726_GWU_018-023.indd 21 4/16/13 10:39 AM
around the object without damaging it. (Rigid-link mechanisms can be made “smart,” too, but the added sensors and controls to do that can complicate the system and make it more prone to failures.) The same idea, only scaled down, can apply to robotic surgery—using the flex- ible arm to prevent inadvertent damage to organs, for instance, from sudden movements. I’m working with surgeons at GW on developing designs for its use in colorectal surgeries. Another project is called STORM, for Self-Configurable and Transformable Omni-Directional Robotic Modules. These are robotic pieces that can assemble into a larger robot and later disassemble. [For example, three small modules could link together to create one robot that looks similar to the hybrid mobile robot on Pg. 21.] If you want to tackle a rubble pile you could send a fixed-configuration robot, but what happens if it doesn’t fit or encounters a fence? It’s stuck. Mission failed. This robot, using a vision system or other sensors, would see it can’t fit and then start disassembling. The robotic pieces would swarm inside the void and communicate with each other wirelessly. If one finds a person stuck under a concrete slab, it can call for the others and connect into a larger configuration capable of handling larger payloads. Once the task is performed, they disassemble and find Employing the brawn of a ready-made PR2 robot, nicknamd Pepe, Dr. Drumwright's lab is focusing on something else to do. the brainpower needed to make robots more useful in the home, particularly for elder care.
It is tele-operated now but we’re
working on another DARPA project to significant amount of force and acts without feedback on an object—or a person—it’s R obotics research is very multi-disci- plinary, so I’ve created a group that reflects that. Having the ability to do all do those things autonomously. We’re going to do more damage than good. the disciplines in-house while the students using a vision-based navigation system We’re working on continuum mecha- interact and learn from one another, I we developed and other sensors—like a nisms for these mobile platforms. Typical think that’s unique. camera in the palm of the hand, stereo robot arms are made up of articulated I like to see the big picture. You can vision for depth perception, laser scan- rigid links. That would be replaced by an have a skeleton but, without the tissues, the ners for accurate measurements of what’s arm that is continuously flexible—it’s muscles, the senses, and the brain, it can’t in front of the robot—that allow you to a flexible unit with multiple segments do anything. I see robotics in a similar way: plan the motions needed, for example, in that allow it to bend like a snake or an the mechanism, the motors and actuators, order to climb difficult obstacles. elephant trunk. the sensors, the wire connections, and the Search-and-rescue robots also have to be If the robot exerts too much force, processor work together to bring the robotic able to act and react safely. If a robot has a the continuum arm will inherently bend system to life.
22 GW RESEARCH • Spring 2013
40726_GWU_018-023.indd 22 4/17/13 7:31 AM
Evan there one thing we can do here that can We’ve also been working on efficiency Drumwright indicate we can do 10 other things? in control. One of my Ph.D. students was One important task we’re focusing on just in Italy for three months working Department of Computer Science is opening doors. It’s something that all with a quadrupedal robot to test our Sample projects: manipulation, robots will need to do and there’s been method of inverse dynamics control. efficient control, and balance limited success. If you have a particular Basically, the idea behind that is accu- kind of door—one that the robot has rately determining what forces need to We don’t have many robots in the home. been specially programmed to open—it be applied to the robot’s motors to make We’ve got the Roomba, and the same can open it. It’s not anywhere as smooth something happen, without using too company also has a similar thing called as a human, but they can do it. much energy. If I’m picking up that vase Scooba for cleaning floors, and we’ve But think of all the different kinds of and I know how much water is in it, how got some toy robots. But we really don’t doors we encounter. You’ve got doors much force do I need to apply to pick it have anything that can help us that where it’s not clear even to humans up? If I’ve got a powerlifter picking up a much in the home. That’s one of the whether you push them or pull them. tiny little vase with all his strength, it’s things that I’m really interested in. I was just in a Starbucks in Japan and not going to be good for energy usage or I use a PR2 [manufactured by Willow thought a door was locked, but it turned for accuracy. Garage], which is a very capable robot. out to be a sliding door. It turns out to be a really hard problem The creator wanted a robot that could What I’d like for the robot to be able to solve because determining what forces make him breakfast. My vision is a to do is open any door autonomously. It’s to apply to the robot’s motors is coupled robot to help me get around when I’m a problem that’s part perception (Have I to determining the friction forces that are old and have just a couple marbles seen this door before?), part modeling (How at the robot’s feet. The two have to be rolling around in my head—to help do I think the mechanism behind the door oper- figured out simultaneously, and our care for me so I don’t put that burden ates?), and part mechanical. method does that. on family or have to pay people who We want the robot to be able to learn This could help improve the accuracy of don’t have a real incentive to do a great from doors that it’s opened before. And movement for any robot that comes into job taking care of me. if the robot hasn’t seen a type of door contact with its environment—so, basically It may not necessarily be doing health before, we want it to push and poke until everything except flying robots—including care, but more like changing a light bulb it figures out how the mechanism works. movement on sticky surfaces, like asphalt, or doing some mild cleaning—you’d be and slick surfaces, like ice and metal. GW amazed at the number of things that elderly people need help with. W e’ve also been working on balance. If somebody gave me skis and goggles and a hat—or in the context of
O ne of the things I’ve been working
on has been manipulation, trying to get the robot to do things dexterously, like a helping the elderly, groceries—how would I carry them? How would a robot do that? One of the ways we did this was through human would do, and to do them at higher force feedback in the robot’s arms. If you’re speeds and in dynamic environments. going to pick up an opaque vase you have There’s a famous video where a PR2 an idea of how much that vase weighs. If folds towels. It takes, I want to say, about somebody had unexpectedly put water into 30 minutes for the robot to fold one towel. it, then all of the sudden you’d get thrown People see the promise in what the robot’s for a loop. Where you determined that doing but it’s so slow they start to joke: things went awry, that was because of force OK, well a $400,000 robot folds one towel in 30 feedback in your arms. minutes, so this will be practical … when? We used that to balance objects on the The other problem is that the PR2’s arms and we ended up being able robots are always doing these things in to do it really well, whereas in the past controlled environments. PR2 stands you would have used the robot’s vision, for Personal Robot 2. It’s meant to be in which is just not as well suited for it. your house doing things alongside you. We don’t want to focus on getting the Robots opening doors autonomously is one task Dr. Drumwright is trying to crack. Faced with a robot to do one thing at a time and creating new type of door, “we want it to push and poke a library of tasks. More importantly: Is until it figures out how the mechanism works.”